- Devel - libvirt List Archives

[libvirt] [RFC] Adopting 'Tested-by' tag (and probably other tags)?
by Kashyap Chamarthy 26 Apr '17

26 Apr '17

Hi folks, I was wondering if the upstream would be receptive to adding a 'Tested-by' tag for patches that had someone give tested feedback on the list. Personally, what I consider a bar for giving a 'Tested-by' is, when someone: - Applies a patch or a series (following the iterations as needed, over time) locally, compile them, spend time understanding the functional change in question, and its implications.) - Do a concrete test (w/ either a Python API or a `virsh`-based command-line) that exercises the said code path. - *Post* the above test procedure / and necessary outputs as evidence to the list, in reasonable detail. (Not a: "yeah, I tested and it works" -- this isn't a 'rule'; it can have exceptions) >From a quick glance, in its 10-year history, libvirt upstream had about a mere 21 entries of 'Tested-by' tags: $ git log | grep Tested-by | wc -l 21 I bring this up because, when someone spends effort following (sometimes over weeks / months) a patch (or a series) from the list, gives reasonably detailed test feedback, in my books, it would be fair to acknowledge it in the Git. And it will encourage them to spend time in future. --- I realize that if it's not automated (via Git hooks or similar), it can become "lossy", i.e. if Joe posts v1 of a patch, you give a 'Tested-by', then there are two scenarios that immediately spring to mind: (1) Joe respins a v2 to make some corrections, adds your 'Tested-by' tag, and whoever applies the patch picks it up -- all good. (b) However, if a v2 was _not_ necessary, then whoever is applying the patch / series must remember to add the tag -- "lossy". Thoughts / remarks / rotten tomatoes welcome. -- /kashyap

5 9

[libvirt] [PATCH v2 00/10] Add mdev reporting capability to the nodedev driver
by Erik Skultety 26 Apr '17

26 Apr '17

since v1: - dropped the <description> element from the parent device nested capability - added missing RNG schema and tests - updated the documentation to describe the MDEV elements in both the parent and the child So, regarding the discussion about the presence of <description> element in the device XML, I dropped it for v2. There were points that we could continue debating about, but one of the major points against it is that libvirt should understand the information it's gathering and then exposing it in a structured way, which in this case wasn't the case, which wouldn't make the management layer's job any easier in any way, since they'd have to parse it either way (from libvirt or on their own). So, until this is resolved among vendors so that we can apply some kind of device classification (no userspace DB), management will have to simply read the description directly from sysfs. As part of a private discussion, another interesting point has been raised - virsh nodedev-list does support '--cap' option which filters only devices matching the pattern. However, with SRIOV, NPIV, and MDEV we have nested capabilities which are kept private and cannot be filtered by. That might be a nice feature for users: 1) OK, let's see what MDEVs are there on the host virsh nodedev-list --cap="mdev-<child|whatever>" 2) I'd like to create some on my own, so which physical devices do support MDEV? virsh nodedev-list --cap='mdev-<parent|whatever>' 3) create the device on that physical device The same goes for SRIOV, where we have virt_functions in parent and phys_function in child and NPIV (I don't remember what it was here), I didn't check in depth, but I think it should be possible for us to do and it's a valid use case where users don't have to parse dumpxml's output for information about the parent's nested capabilities. Erik Erik Skultety (10): nodedev: Make use of the compile-time missing enum in switch error conf: nodedev: Split virNodeDeviceDefFormat into more functions nodedev: udevProcessPCI: Drop syspath variable docs: Utilize our XSLT list generating template more nodedev: conf: Split PCI sub-capability parsing to a separate method nodedev: Introduce the mdev capability to the nodedev driver structure nodedev: Introduce the mdev capability to a PCI parent device nodedev: Introduce mdev capability for child devices docs: Provide a nodedev driver stub documentation docs: Document the mediated devices within the nodedev driver docs/drivers.html.in | 6 +- docs/drvnodedev.html.in | 344 +++++++++ docs/remote.html.in | 106 +-- docs/schemas/nodedev.rng | 41 ++ docs/storage.html.in | 62 +- include/libvirt/libvirt-nodedev.h | 1 + src/conf/node_device_conf.c | 812 +++++++++++++-------- src/conf/node_device_conf.h | 21 +- src/conf/virnodedeviceobj.c | 3 +- src/libvirt-nodedev.c | 1 + src/libvirt_private.syms | 1 + src/node_device/node_device_driver.c | 1 + src/node_device/node_device_udev.c | 242 +++++- .../mdev_3627463d_b7f0_4fea_b468_f1da537d301b.xml | 8 + tests/nodedevschemadata/pci_0000_02_10_7_mdev.xml | 27 + tests/nodedevxml2xmltest.c | 1 + tools/virsh-nodedev.c | 2 + 17 files changed, 1223 insertions(+), 456 deletions(-) create mode 100644 docs/drvnodedev.html.in create mode 100644 tests/nodedevschemadata/mdev_3627463d_b7f0_4fea_b468_f1da537d301b.xml create mode 100644 tests/nodedevschemadata/pci_0000_02_10_7_mdev.xml -- 2.12.2

3 23

Re: [libvirt] How to get VM memory usage
by Eric Blake 26 Apr '17

26 Apr '17

[moderator note: I'm forwarding a stripped down version of the original mail which was rejected in the moderator queue. I stripped the .jpg attachments, which are inappropriate for a technical list. Either paste the text contents (rather than a screenshot) of the window you are describing, or host your images elsewhere and have your list email merely give a URL of the externally-hosted images] > ForwardedMessage.eml > > Subject: > Re: Re: [libvirt] How to get VM memory usage > From: > "wangyufeng(a)cloudguarding.com" <wangyufeng(a)cloudguarding.com> > Date: > 04/26/2017 01:21 AM > > To: > "Daniel P. Berrange" <berrange(a)redhat.com> > CC: > libvir-list <libvir-list(a)redhat.com> > > > Hi Daniel, > > I found method MemoryStats in domain.go: > func (d *Domain) MemoryStats(nrStats uint32, flags uint32) ([]DomainMemoryStat, error) { > ptr := make([]C.virDomainMemoryStatStruct, nrStats) > > result := C.virDomainMemoryStats( > d.ptr, (C.virDomainMemoryStatPtr)(unsafe.Pointer(&ptr[0])), > C.uint(nrStats), C.uint(flags)) > > if result == -1 { > return []DomainMemoryStat{}, GetLastError() > } > > out := make([]DomainMemoryStat, result) > for i := 0; i < int(result); i++ { > out = append(out, DomainMemoryStat{ > Tag: int32(ptr[i].tag), > Val: uint64(ptr[i].val), > }) > } > return out, nil > } > > Do you mean this method can get used memory info? > While I invoke the method, which can only get values of tag 0 and tag 6: > > And the balloon driver have been activated on the monitored vm: > > Below is my code, could you tell me what's wrong with my code? > > Thank you very much. > > ---------------------------------------------------------------------------------------------------------- > On Fri, Apr 21, 2017 at 09:58:12AM +0800, wangyufeng(a)cloudguarding.com wrote: >> Hi Daniel, >> >> In method 'GetInfo', we can only get memory and max memory. However I >> need to get used memory for calculating memory usage. The virsh command >> 'dommemstat' can obtain used memory, is there a method in libvirt-go to >> obtain used memory? > > That command uses the virDomainMemoryStats() API call > > > Regards, > Daniel > -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange > :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: > https://entangle-photo.org -o- https://www.instagram.com/dberrange :| -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org

1 0

[libvirt] [PATCH v3 0/5] Alter the clearing of device for new partition table
by John Ferlan 26 Apr '17

26 Apr '17

Well kind of a v2/v3 type mashup - essentially reworked the portion of the series: https://www.redhat.com/archives/libvir-list/2017-April/msg00402.html that dealt with the zeroing of the header of the device path. The changes now will allow both the head and tail of the device path to be cleared rather than just the header of the file which was ACK'd for at least the logical backend. Additionally, bump the size of the clear/zero from 2048KB to 1MB and use it for both the logical and disk backends. John Ferlan (5): storage: Modify storageBackendWipeLocal to allow zero from end of device storage: Introduce virStorageBackendZeroPartitionTable logical: Use virStorageBackendZeroPartitionTable logical: Increase the size of the data to wipe disk: Use virStorageBackendZeroPartitionTable src/storage/storage_backend_disk.c | 6 ++- src/storage/storage_backend_logical.c | 44 +--------------------- src/storage/storage_util.c | 70 ++++++++++++++++++++++++++++------- src/storage/storage_util.h | 4 ++ 4 files changed, 67 insertions(+), 57 deletions(-) -- 2.9.3

2 10

[libvirt] [PATCH] Update keycodemapdb submodule for python2 compat fixes
by Daniel P. Berrange 26 Apr '17

26 Apr '17

There were a few bugs in keycodemap tool that broke it when run on python circa 2.7.5 or older, which affected RHEL builds. Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- Pushed as a CI build fix for rhel6/7 src/keycodemapdb | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/keycodemapdb b/src/keycodemapdb index bd754b8..7bf5710 160000 --- a/src/keycodemapdb +++ b/src/keycodemapdb @@ -1 +1 @@ -Subproject commit bd754b861e7de5c7919aa249a975913cb207b85d +Subproject commit 7bf5710b22aa8d58b7eeaaf3dc6960c26cade4f0 -- 2.9.3

1 0

[libvirt] [PATCH v2 00/13] Multiple cleanups within nwfilterobj and nwfilter drivers
by John Ferlan 26 Apr '17

26 Apr '17

v1: https://www.redhat.com/archives/libvir-list/2017-April/msg01074.html Changes since v1: * Patch 2 -> Do not add the local @def when just dereference obj->def * Patch 13 & 15 dropped Most patches were ACK'd, but patch 2's impact was felt in a few other patches so I just resent the whole thing. In particular patch 5 for nwfilterUndefine and nwfilterGetXMLDesc as well as patch 7 for the _virNWFilterObjListDefLoopDetect call w/ _virNWFilterObjListDefLoopDetect John Ferlan (13): nwfilter: Use consistent naming for variables nwfilter: Use virNWFilterDefPtr rather than deref virNWFilterObjPtr nwfilter: Remove unused 'active' in virNWFilterObj nwfilter: Convert wantRemoved to bool nwfilter: Make _virNWFilterObjPtr private nwfilter: Introduce virNWFilterObjNew nwfilter: Rename some virNWFilterObj* API's nwfilter: Make _virNWFilterObjList private nwfilter: Make a common UUID lookup function from driver nwfilter: Replace virNWFilterConfigFile with virFileBuildPath nwfilter: Replace virNWFilterSaveDef with virNWFilterSaveConfig nwfilter: Move creation of configDir to driver initialization nwfilter: Move save of config until after successful assign src/conf/nwfilter_conf.c | 57 +------ src/conf/nwfilter_conf.h | 8 - src/conf/virnwfilterobj.c | 270 +++++++++++++++++++++------------ src/conf/virnwfilterobj.h | 70 ++++----- src/libvirt_private.syms | 20 ++- src/nwfilter/nwfilter_driver.c | 138 ++++++++++------- src/nwfilter/nwfilter_gentech_driver.c | 42 ++--- 7 files changed, 320 insertions(+), 285 deletions(-) -- 2.9.3 FWIW: The diffs between this series and the previous taking away patches 13/15: diff --git a/src/conf/virnwfilterobj.c b/src/conf/virnwfilterobj.c index 62206b8..8c9da13 100644 --- a/src/conf/virnwfilterobj.c +++ b/src/conf/virnwfilterobj.c @@ -92,16 +92,11 @@ virNWFilterObjWantRemoved(virNWFilterObjPtr obj) static void virNWFilterObjFree(virNWFilterObjPtr obj) { - virNWFilterDefPtr def; - virNWFilterDefPtr newDef; - if (!obj) return; - def = obj->def; - newDef = obj->newDef; - virNWFilterDefFree(def); - virNWFilterDefFree(newDef); + virNWFilterDefFree(obj->def); + virNWFilterDefFree(obj->newDef); virMutexDestroy(&obj->lock); @@ -204,7 +199,6 @@ _virNWFilterObjListDefLoopDetect(virNWFilterObjListPtr nwfilters, size_t i; virNWFilterEntryPtr entry; virNWFilterObjPtr obj; - virNWFilterDefPtr objdef; if (!def) return 0; @@ -221,8 +215,7 @@ _virNWFilterObjListDefLoopDetect(virNWFilterObjListPtr nwfilters, obj = virNWFilterObjListFindByName(nwfilters, entry->include->filterref); if (obj) { - objdef = obj->def; - rc = _virNWFilterObjListDefLoopDetect(nwfilters, objdef, + rc = _virNWFilterObjListDefLoopDetect(nwfilters, obj->def, filtername); virNWFilterObjUnlock(obj); if (rc < 0) @@ -386,13 +379,11 @@ virNWFilterObjListNumOfNWFilters(virNWFilterObjListPtr nwfilters, { size_t i; int nfilters = 0; - virNWFilterDefPtr def; for (i = 0; i < nwfilters->count; i++) { virNWFilterObjPtr obj = nwfilters->objs[i]; virNWFilterObjLock(obj); - def = obj->def; - if (!aclfilter || aclfilter(conn, def)) + if (!aclfilter || aclfilter(conn, obj->def)) nfilters++; virNWFilterObjUnlock(obj); }

2 14

[libvirt] Unbounded client streams
by Michal Privoznik 26 Apr '17

26 Apr '17

Dear list, as you might have seen, I've been playing with our streams lately. One of the things I wanted to try was how my sparse streams deal with vol-download and slow writer (e.g. slow disk). Note to whomever wants to try that out: blkio,throttle.write_* is no good for this. So I went the old way: diff --git i/tools/virsh-util.c w/tools/virsh-util.c index 4b86e29..8f5738d 100644 --- i/tools/virsh-util.c +++ w/tools/virsh-util.c @@ -149,6 +149,8 @@ virshStreamSink(virStreamPtr st ATTRIBUTE_UNUSED, { int *fd = opaque; + sleep(60); + return safewrite(*fd, bytes, nbytes); } And found out something ugly: the memory footprint of virsh just keeps growing as vol-download progresses. I've tracked down the problem to: 1) The client event loop (virNetClientIOEventLoop) sees some incoming data, so it reads it (virNetClientIOHandleInput -> virNetClientCallDispatch) and queues the packet (virNetClientCallDispatchStream) regardless of the length of already queued packets 2) Since the server sees client has read all the data, it sends even more data. And the queue in step 1) just grows and grows. Ideally, the packets are taken out from the queue by virStreamRecv (which boils down to virNetClientStreamRecvPacket), but that is not on the schedule for the next minute. Frankly, I don't have any idea how to fix this. We can stop reading the data (i.e. not set POLLIN on the client <-> server socket) if the queue reaches certain length. BUT, this would mean no other call is processed. Not even keepalive - and the connection would break. The other option is to pretend I've never send this e-mail and you've read nothing O:-) Michal

4 9

[libvirt] [PATCHv2 0/6] Support more options for intel-iommu
by Ján Tomko 26 Apr '17

26 Apr '17

For https://bugzilla.redhat.com/show_bug.cgi?id=1427005 v2: * use != VIR_TRISTATE_SWITCH_ABSENT more often * use symlinks for identical XMLs * move capability checks in qemuBuildIOMMUCommandLine to the beginning, to be shared with qemuBuildIOMMUCommandLine Documentation changes reflecting current QEMU implementation limits are not included. Ján Tomko (6): conf: add <irqchip mode> to <features> qemu: format kernel_irqchip on the command line conf: add <driver intremap> to <iommu> qemu: format intremap= on intel-iommu command line conf: add caching attribute to iommu device qemu: format caching-mode on iommu command line docs/formatdomain.html.in | 41 +++++++++- docs/schemas/domaincommon.rng | 30 ++++++++ src/conf/domain_conf.c | 90 ++++++++++++++++++++-- src/conf/domain_conf.h | 14 ++++ src/libvirt_private.syms | 1 + src/qemu/qemu_capabilities.c | 12 +++ src/qemu/qemu_capabilities.h | 3 + src/qemu/qemu_command.c | 40 ++++++++++ tests/qemucapabilitiesdata/caps_1.5.3.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_1.6.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_1.7.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_2.1.1.x86_64.xml | 1 + .../qemucapabilitiesdata/caps_2.4.0.x86_64.replies | 22 ++++-- tests/qemucapabilitiesdata/caps_2.4.0.x86_64.xml | 1 + .../qemucapabilitiesdata/caps_2.5.0.x86_64.replies | 24 ++++-- tests/qemucapabilitiesdata/caps_2.5.0.x86_64.xml | 1 + .../caps_2.6.0-gicv2.aarch64.xml | 1 + .../caps_2.6.0-gicv3.aarch64.xml | 1 + tests/qemucapabilitiesdata/caps_2.6.0.ppc64le.xml | 1 + .../qemucapabilitiesdata/caps_2.6.0.x86_64.replies | 24 ++++-- tests/qemucapabilitiesdata/caps_2.6.0.x86_64.xml | 1 + tests/qemucapabilitiesdata/caps_2.7.0.s390x.xml | 1 + .../qemucapabilitiesdata/caps_2.7.0.x86_64.replies | 28 +++++-- tests/qemucapabilitiesdata/caps_2.7.0.x86_64.xml | 2 + tests/qemucapabilitiesdata/caps_2.8.0.s390x.xml | 1 + .../qemucapabilitiesdata/caps_2.8.0.x86_64.replies | 37 +++++++-- tests/qemucapabilitiesdata/caps_2.8.0.x86_64.xml | 2 + .../qemucapabilitiesdata/caps_2.9.0.x86_64.replies | 49 +++++++++--- tests/qemucapabilitiesdata/caps_2.9.0.x86_64.xml | 3 + .../qemuxml2argv-intel-iommu-caching.args | 19 +++++ .../qemuxml2argv-intel-iommu-caching.xml | 28 +++++++ .../qemuxml2argv-intel-iommu-irqchip.args | 19 +++++ .../qemuxml2argv-intel-iommu-irqchip.xml | 31 ++++++++ tests/qemuxml2argvtest.c | 10 +++ .../qemuxml2xmlout-intel-iommu-caching.xml | 1 + .../qemuxml2xmlout-intel-iommu-irqchip.xml | 1 + tests/qemuxml2xmltest.c | 2 + 37 files changed, 496 insertions(+), 49 deletions(-) create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-intel-iommu-caching.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-intel-iommu-caching.xml create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-intel-iommu-irqchip.args create mode 100644 tests/qemuxml2argvdata/qemuxml2argv-intel-iommu-irqchip.xml create mode 120000 tests/qemuxml2xmloutdata/qemuxml2xmlout-intel-iommu-caching.xml create mode 120000 tests/qemuxml2xmloutdata/qemuxml2xmlout-intel-iommu-irqchip.xml -- 2.10.2

2 14

[libvirt] [PATCH] Support CPU less NUMA node.
by Shaohe Feng 26 Apr '17

26 Apr '17

Some platform NUMA node does not include cpu, such as Phi. This patch support CPU less NUMA node. But there must be one CPU cell node for a guest. Also must assign the host nodeset for this guest cell node. please enable numactl with "--with-numactl" for libvirt config. Test this patch: 1. define a numa cell without "cpus", such as cell 1. <numatune> <memnode cellid='1' mode='strict' nodeset='0'/> </numatune> <cpu> <numa> <cell id='0' cpus='0-1' memory='512000' unit='KiB'/> <cell id='1' memory='512000' unit='KiB'/> </numa> </cpu> libvirt can edit and start the VM successfully. 2. define a numa cell without "cpus", without nodeset. <cpu> <numa> <cell id='0' cpus='0-1' memory='512000' unit='KiB'/> <cell id='1' memory='512000' unit='KiB'/> </numa> </cpu> This case, libvirt will failed to edit the CPUS. --- src/conf/domain_conf.c | 4 +++ src/conf/numa_conf.c | 89 ++++++++++++++++++++++++++++++++++--------------- src/conf/numa_conf.h | 2 ++ src/qemu/qemu_command.c | 12 ++++--- src/util/virbitmap.c | 10 ++++-- 5 files changed, 83 insertions(+), 34 deletions(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 705deb3..d6e5489 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -17330,6 +17330,10 @@ virDomainDefParseXML(xmlDocPtr xml, ctxt) < 0) goto error; + if (virDomainNumaCPULessCheck(def->numa) < 0){ + goto error; + } + if (virDomainNumatuneHasPlacementAuto(def->numa) && !def->cpumask && !virDomainDefHasVcpuPin(def) && !def->cputune.emulatorpin && diff --git a/src/conf/numa_conf.c b/src/conf/numa_conf.c index bfd3703..1b6f7ff 100644 --- a/src/conf/numa_conf.c +++ b/src/conf/numa_conf.c @@ -742,34 +742,31 @@ virDomainNumaDefCPUParseXML(virDomainNumaPtr def, goto cleanup; } - if (!(tmp = virXMLPropString(nodes[i], "cpus"))) { - virReportError(VIR_ERR_XML_ERROR, "%s", - _("Missing 'cpus' attribute in NUMA cell")); - goto cleanup; - } - - if (virBitmapParse(tmp, &def->mem_nodes[cur_cell].cpumask, - VIR_DOMAIN_CPUMASK_LEN) < 0) - goto cleanup; - - if (virBitmapIsAllClear(def->mem_nodes[cur_cell].cpumask)) { - virReportError(VIR_ERR_CONFIG_UNSUPPORTED, - _("NUMA cell %d has no vCPUs assigned"), cur_cell); - goto cleanup; - } - VIR_FREE(tmp); - - for (j = 0; j < n; j++) { - if (j == cur_cell || !def->mem_nodes[j].cpumask) - continue; + tmp = virXMLPropString(nodes[i], "cpus"); + if (tmp) { + if (virBitmapParse(tmp, &def->mem_nodes[cur_cell].cpumask, + VIR_DOMAIN_CPUMASK_LEN) < 0) + goto cleanup; - if (virBitmapOverlaps(def->mem_nodes[j].cpumask, - def->mem_nodes[cur_cell].cpumask)) { + if (virBitmapIsAllClear(def->mem_nodes[cur_cell].cpumask)) { virReportError(VIR_ERR_CONFIG_UNSUPPORTED, - _("NUMA cells %u and %zu have overlapping vCPU ids"), - cur_cell, j); + _("NUMA cell %d has no vCPUs assigned"), cur_cell); goto cleanup; } + VIR_FREE(tmp); + + for (j = 0; j < n; j++) { + if (j == cur_cell || !def->mem_nodes[j].cpumask) + continue; + + if (virBitmapOverlaps(def->mem_nodes[j].cpumask, + def->mem_nodes[cur_cell].cpumask)) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("NUMA cells %u and %zu have overlapping vCPU ids"), + cur_cell, j); + goto cleanup; + } + } } ctxt->node = nodes[i]; @@ -808,6 +805,7 @@ virDomainNumaDefCPUFormat(virBufferPtr buf, char *cpustr; size_t ncells = virDomainNumaGetNodeCount(def); size_t i; + size_t ncpuless = 0; if (ncells == 0) return 0; @@ -817,12 +815,24 @@ virDomainNumaDefCPUFormat(virBufferPtr buf, for (i = 0; i < ncells; i++) { memAccess = virDomainNumaGetNodeMemoryAccessMode(def, i); - if (!(cpustr = virBitmapFormat(virDomainNumaGetNodeCpumask(def, i)))) - return -1; + cpustr = virBitmapFormat(virDomainNumaGetNodeCpumask(def, i)); virBufferAddLit(buf, "<cell"); virBufferAsprintf(buf, " id='%zu'", i); - virBufferAsprintf(buf, " cpus='%s'", cpustr); + if (cpustr && (*cpustr != 0)) + virBufferAsprintf(buf, " cpus='%s'", cpustr); + else { + /* Note, at present we only allow cpuless numa node with */ + /* nodeset assign. */ + ncpuless++; + if (!virDomainNumatuneNodeSpecified(def, i)) { + VIR_FREE(cpustr); + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("CPUless NUMA node cellid %zu must be " + "specified node set."), i); + return -1; + } + } virBufferAsprintf(buf, " memory='%llu'", virDomainNumaGetNodeMemorySize(def, i)); virBufferAddLit(buf, " unit='KiB'"); @@ -835,6 +845,12 @@ virDomainNumaDefCPUFormat(virBufferPtr buf, virBufferAdjustIndent(buf, -2); virBufferAddLit(buf, "</numa>\n"); + if (ncpuless == ncells) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("At lease one NUMA node should " + "include CPU element.")); + return -1; + } return 0; } @@ -861,6 +877,7 @@ virDomainNumaGetMaxCPUID(virDomainNumaPtr numa) int bit; bit = virBitmapLastSetBit(virDomainNumaGetNodeCpumask(numa, i)); + bit = (bit > 0) ? bit : 0; if (bit > ret) ret = bit; } @@ -973,3 +990,21 @@ virDomainNumaGetMemorySize(virDomainNumaPtr numa) return ret; } + +// NOTE, For some plateforms, there are some node without cpus, such as Phi. +// We just need do some check for these plateforms, we must assigned nodeset +// for CPU less node. +int +virDomainNumaCPULessCheck(virDomainNumaPtr numa) +{ + int ret = 0; + int i = 0; + for (i = 0; i < numa->nmem_nodes; i++){ + if (!numa->mem_nodes[i].cpumask && !numa->mem_nodes[i].nodeset) { + virReportError(VIR_ERR_CONFIG_UNSUPPORTED, + _("Cell: %d is a cpuless numa, must specfiy the nodeset."), i); + return -1; + } + } + return ret; +} diff --git a/src/conf/numa_conf.h b/src/conf/numa_conf.h index b6a5354..1f711e6 100644 --- a/src/conf/numa_conf.h +++ b/src/conf/numa_conf.h @@ -155,5 +155,7 @@ int virDomainNumaDefCPUFormat(virBufferPtr buf, virDomainNumaPtr def); unsigned int virDomainNumaGetCPUCountTotal(virDomainNumaPtr numa); +int virDomainNumaCPULessCheck(virDomainNumaPtr numa); + #endif /* __NUMA_CONF_H__ */ diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index b2e76ca..b07ca66 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -7665,11 +7665,13 @@ qemuBuildNumaArgStr(virQEMUDriverConfigPtr cfg, virCommandAddArg(cmd, "-numa"); virBufferAsprintf(&buf, "node,nodeid=%zu", i); - for (tmpmask = cpumask; tmpmask; tmpmask = next) { - if ((next = strchr(tmpmask, ','))) - *(next++) = '\0'; - virBufferAddLit(&buf, ",cpus="); - virBufferAdd(&buf, tmpmask, -1); + if (*cpumask != 0) { + for (tmpmask = cpumask; tmpmask; tmpmask = next) { + if ((next = strchr(tmpmask, ','))) + *(next++) = '\0'; + virBufferAddLit(&buf, ",cpus="); + virBufferAdd(&buf, tmpmask, -1); + } } if (needBackend) diff --git a/src/util/virbitmap.c b/src/util/virbitmap.c index eac63d9..b2803ce 100644 --- a/src/util/virbitmap.c +++ b/src/util/virbitmap.c @@ -953,6 +953,9 @@ virBitmapLastSetBit(virBitmapPtr bitmap) unsigned long bits; /* If bitmap is empty then there is no set bit */ + if (!bitmap) + return -1; + if (bitmap->map_len == 0) return -1; @@ -1039,9 +1042,12 @@ virBitmapCountBits(virBitmapPtr bitmap) size_t i; size_t ret = 0; - for (i = 0; i < bitmap->map_len; i++) - ret += count_one_bits_l(bitmap->map[i]); + if (bitmap == NULL) + return ret; + for (i = 0; i < bitmap->map_len; i++) { + ret += count_one_bits_l(bitmap->map[i]); + } return ret; } -- 2.7.4

3 4

[libvirt] [PATCH] virsh-domain: update attach-interface to support bridge type
by ZhiPeng Lu 26 Apr '17

26 Apr '17

Adding this feature will allow users to attach a network to linux bridge or openvswitch. The interface can be attached using --brtype=linux or --brtype=openvswitch. If --brtype is not set, the network is attached to linux bridge. Signed-off-by: ZhiPeng Lu <lu.zhipeng(a)zte.com.cn> --- tools/virsh-domain.c | 23 +++++++++++++++++++++-- tools/virsh.pod | 3 +++ 2 files changed, 24 insertions(+), 2 deletions(-) diff --git a/tools/virsh-domain.c b/tools/virsh-domain.c index db8accf..2690689 100644 --- a/tools/virsh-domain.c +++ b/tools/virsh-domain.c @@ -766,6 +766,10 @@ static const vshCmdOptDef opts_attach_interface[] = { .type = VSH_OT_BOOL, .help = N_("libvirt will automatically detach/attach the device from/to host") }, + {.name = "brtype", + .type = VSH_OT_STRING, + .help = N_("Bridge type: linux / openvswitch") + }, {.name = NULL} }; @@ -819,7 +823,8 @@ cmdAttachInterface(vshControl *ctl, const vshCmd *cmd) virDomainPtr dom = NULL; const char *mac = NULL, *target = NULL, *script = NULL, *type = NULL, *source = NULL, *model = NULL, - *inboundStr = NULL, *outboundStr = NULL; + *inboundStr = NULL, *outboundStr = NULL, + *brtype = NULL; virNetDevBandwidthRate inbound, outbound; virDomainNetType typ; int ret; @@ -850,7 +855,8 @@ cmdAttachInterface(vshControl *ctl, const vshCmd *cmd) vshCommandOptStringReq(ctl, cmd, "script", &script) < 0 || vshCommandOptStringReq(ctl, cmd, "model", &model) < 0 || vshCommandOptStringReq(ctl, cmd, "inbound", &inboundStr) < 0 || - vshCommandOptStringReq(ctl, cmd, "outbound", &outboundStr) < 0) + vshCommandOptStringReq(ctl, cmd, "outbound", &outboundStr) < 0 || + vshCommandOptStringReq(ctl, cmd, "brtype", &brtype) < 0) goto cleanup; /* check interface type */ @@ -946,6 +952,19 @@ cmdAttachInterface(vshControl *ctl, const vshCmd *cmd) if (model != NULL) virBufferAsprintf(&buf, "<model type='%s'/>\n", model); + if ((brtype != NULL)) { + if (STRNEQ(brtype , "linux") && STRNEQ(brtype , "openvswitch")) { + vshError(ctl, _("bridge type is incorrect, should be linux " + "or openvswitch")); + goto cleanup; + } + if (typ != VIR_DOMAIN_NET_TYPE_BRIDGE) { + vshError(ctl, _("No support for %s with bridge type"), type); + goto cleanup; + } + if (STREQ(brtype , "openvswitch")) + virBufferAsprintf(&buf, " <virtualport type='openvswitch'/>\n"); + } if (inboundStr || outboundStr) { virBufferAddLit(&buf, "<bandwidth>\n"); virBufferAdjustIndent(&buf, 2); diff --git a/tools/virsh.pod b/tools/virsh.pod index e16f62f..3cd0490 100644 --- a/tools/virsh.pod +++ b/tools/virsh.pod @@ -2935,6 +2935,9 @@ B<--managed> is usable only for I<hostdev> type and tells libvirt that the interface should be managed, which means detached and reattached from/to the host by libvirt. +B<--brtype> is usable only for I<bridge> type and tells libvirt +that the interface attached to bridge type (linux bridge or openvswitch). + If B<--print-xml> is specified, then the XML of the interface that would be attached is printed instead. -- 1.8.3.1

2 1

[libvirt] [PATCH 00/15] Multiple cleanups within nwfilterobj and nwfilter drivers
by John Ferlan 25 Apr '17

25 Apr '17

More adjustments in preparation for having virobject code handle the bulk of the object management code. I know it's a lot of patches, but reducing the amount of change overall has the opposite effect of increasing the number patches to show the work flow. Should make the "next" steps easier to review, but a couple of these will be tedious. John Ferlan (15): nwfilter: Use consistent naming for variables nwfilter: Use virNWFilterDefPtr rather than deref virNWFilterObjPtr nwfilter: Remove unused 'active' in virNWFilterObj nwfilter: Convert wantRemoved to bool nwfilter: Make _virNWFilterObjPtr private nwfilter: Introduce virNWFilterObjNew nwfilter: Rename some virNWFilterObj* API's nwfilter: Make _virNWFilterObjList private nwfilter: Make a common UUID lookup function from driver nwfilter: Replace virNWFilterConfigFile with virFileBuildPath nwfilter: Replace virNWFilterSaveDef with virNWFilterSaveConfig nwfilter: Move creation of configDir to driver initialization nwfilter: Move configFile name processing into driver nwfilter: Move save of config until after successful assign nwfilter: Move and rename virNWFilterSaveConfig src/conf/nwfilter_conf.c | 107 +----------- src/conf/nwfilter_conf.h | 16 +- src/conf/virnwfilterobj.c | 308 ++++++++++++++++++++++----------- src/conf/virnwfilterobj.h | 74 ++++---- src/libvirt_private.syms | 21 ++- src/nwfilter/nwfilter_driver.c | 148 ++++++++++------ src/nwfilter/nwfilter_gentech_driver.c | 42 +++-- 7 files changed, 377 insertions(+), 339 deletions(-) -- 2.9.3

2 33

[libvirt] [PATCH v3 0/2] Use the shared keycodemap database
by Daniel P. Berrange 25 Apr '17

25 Apr '17

Follow up to: v1: https://www.redhat.com/archives/libvir-list/2017-March/msg00070.html v2: https://www.redhat.com/archives/libvir-list/2017-March/msg00136.html Changed in v3: - Rebase onto new autogen.sh - Use VIR_KEYMAP_ENTRY_MAX constant more - Change style of make vars - Fix & verify distcheck Changed in v2: - Fixed make syntax-check errors - Auto-generate man pages listing value key codes for use with the virDomainSendKey API / virsh send-key command Daniel P. Berrange (2): util: switch over to use keycodemapdb GIT submodule Add ability to generate man page describing key code names & values .gitignore | 3 +- .gitmodules | 3 + libvirt.spec.in | 1 + mingw-libvirt.spec.in | 2 + src/Makefile.am | 86 ++++++-- src/keycodemapdb | 1 + src/util/keymaps.csv | 464 ------------------------------------------ src/util/virkeycode-mapgen.py | 97 --------- src/util/virkeycode.c | 94 ++++----- tests/virkeycodetest.c | 5 + tools/virsh.pod | 20 ++ 11 files changed, 153 insertions(+), 623 deletions(-) create mode 160000 src/keycodemapdb delete mode 100644 src/util/keymaps.csv delete mode 100755 src/util/virkeycode-mapgen.py -- 2.9.3

2 5

[libvirt] [PATCH 00/14] Multiple cleanups within secretobj and secret_driver
by John Ferlan 25 Apr '17

25 Apr '17

More adjustments in preparation for having virobject code handle the bulk of the object management code. I know it's a lot of patches, but most should be relatively simple - especially the first 10... The last 4 set up for a "common" way to handle the config directory space as well and could be a bit more controversial. Perhaps the most being patch 12 which attempts to perform the mkdir of the configDir during initialization rather than first "define" of secret. Although someone could also pick out their own most favorite/controversial and surprise me. John Ferlan (14): secret: Need to set data->error on VIR_ALLOC_N failure secret: Convert virsecretobjs.h to use "newer" formatting style secret: Make some virSecretObj* functions static secret: Have virSecretObjNew return locked object secret: Use consistent naming for variables secret: Use virSecretDefPtr rather than deref from virSecretObjPtr secret: Move virSecretObjListGetUUIDs secret: Change variable names for list traversals secret: Split apart NumOfSecrets and GetUUIDs callback function secret: Combine virSecretObjListAdd with Locked function secret: Alter FindByUUID to expect the formatted uuidstr secret: Generate configDir during driver initialization secret: Alter configFile/base64File mgmt secret: Clean up virSecretObjListExport logic src/conf/virsecretobj.c | 557 +++++++++++++++++++++++---------------------- src/conf/virsecretobj.h | 120 +++++----- src/secret/secret_driver.c | 178 ++++++++------- 3 files changed, 442 insertions(+), 413 deletions(-) -- 2.9.3

2 51

[libvirt] [PATCH v2 0/2] Augment debug messages.
by Prerna Saxena 25 Apr '17

25 Apr '17

This is a v2 of the previous set of debug enhancements posted at : https://www.redhat.com/archives/libvir-list/2017-March/msg00959.html Changelog: --------- 1) Patch 1/3 : virNetSocketGetFD() is reverted to its old state, and only new DEBUG messages are introduced. 2) Patch 2/3 : Dropped, per list suggestions. 3) Patch 3/3 : Same as v1 Prerna Saxena (2): Debug: Add DEBUG messages for when a client has opened/closed connection. Debug: Report VM errors more concisely. src/qemu/qemu_process.c | 4 ++-- src/rpc/virnetserverclient.c | 10 +++++++++- 2 files changed, 11 insertions(+), 3 deletions(-) -- 1.8.1.2

3 4

[libvirt] [PATCH v4 1/2] daemon: Fix domain name leak in error path
by Wang King 25 Apr '17

25 Apr '17

Domain name duplicated in make_nonnull_domain, but not freed when virTypedParamsSerialize return negative. --- daemon/remote.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/daemon/remote.c b/daemon/remote.c index 1610fea..a8c21fd 100644 --- a/daemon/remote.c +++ b/daemon/remote.c @@ -1066,8 +1066,10 @@ remoteRelayDomainEventTunable(virConnectPtr conn, if (virTypedParamsSerialize(params, nparams, (virTypedParameterRemotePtr *) &data.params.params_val, &data.params.params_len, - VIR_TYPED_PARAM_STRING_OKAY) < 0) + VIR_TYPED_PARAM_STRING_OKAY) < 0) { + VIR_FREE(data.dom.name); return -1; + } remoteDispatchObjectEventSend(callback->client, remoteProgram, REMOTE_PROC_DOMAIN_EVENT_CALLBACK_TUNABLE, @@ -1206,8 +1208,10 @@ remoteRelayDomainEventJobCompleted(virConnectPtr conn, if (virTypedParamsSerialize(params, nparams, (virTypedParameterRemotePtr *) &data.params.params_val, &data.params.params_len, - VIR_TYPED_PARAM_STRING_OKAY) < 0) + VIR_TYPED_PARAM_STRING_OKAY) < 0) { + VIR_FREE(data.dom.name); return -1; + } remoteDispatchObjectEventSend(callback->client, remoteProgram, REMOTE_PROC_DOMAIN_EVENT_CALLBACK_JOB_COMPLETED, -- 2.8.3

2 3

[libvirt] [PATCH] qemu: fix argument of virDomainNetGetActualDirectMode
by ZhiPeng Lu 25 Apr '17

25 Apr '17

it should be a comparison of modes between new and old devices. So the argument of the second virDomainNetGetActualDirectMode should be newdev. Signed-off-by: ZhiPeng Lu <lu.zhipeng(a)zte.com.cn> --- src/qemu/qemu_hotplug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/qemu/qemu_hotplug.c b/src/qemu/qemu_hotplug.c index 120bcdc..eec99af 100644 --- a/src/qemu/qemu_hotplug.c +++ b/src/qemu/qemu_hotplug.c @@ -3273,7 +3273,7 @@ qemuDomainChangeNet(virQEMUDriverPtr driver, if (STRNEQ_NULLABLE(virDomainNetGetActualDirectDev(olddev), virDomainNetGetActualDirectDev(newdev)) || - virDomainNetGetActualDirectMode(olddev) != virDomainNetGetActualDirectMode(olddev) || + virDomainNetGetActualDirectMode(olddev) != virDomainNetGetActualDirectMode(newdev) || !virNetDevVPortProfileEqual(virDomainNetGetActualVirtPortProfile(olddev), virDomainNetGetActualVirtPortProfile(newdev)) || !virNetDevVlanEqual(virDomainNetGetActualVlan(olddev), -- 1.8.3.1

2 1

[libvirt] [PATCH v2] autogen.sh: Improve and generalize
by Andrea Bolognani 25 Apr '17

25 Apr '17

The goal is twofold: firstly, we want to extend the script so that it can deal with more than a single git submodule, and secondly we'd like to reduce the amount of duplicated code. Moreover, since we're making heavy changes to the code anyway, we might as well make sure it follows a somewhat consistent coding style too. To reduce code duplication, we introduce a new --dry-run option, which can be used by third parties to figure out whether calling autogen.sh is necessary or not: this allows us to get rid of the reimplementation of part of the logic in cfg.mk and guarantee they'll never get out of sync. Other changes include: making dirty submodules checking and cleaning entirely independent of other operations; removing the use of 'set -e' and handling errors explicitly instead; better parsing of command line arguments. --- Changes from [v1]: * drop the generic submodule update framework and simply update all submodules unconditionally, *then* perform any required gnulib-specific step; * handle "autogen not required" and "dry run error" differently in cfg.mk; * improve comments. [v1] https://www.redhat.com/archives/libvir-list/2017-April/msg00816.html autogen.sh | 253 ++++++++++++++++++++++++++++++++++++++++--------------------- cfg.mk | 53 ++++++------- 2 files changed, 193 insertions(+), 113 deletions(-) diff --git a/autogen.sh b/autogen.sh index d1c319d..d5d836a 100755 --- a/autogen.sh +++ b/autogen.sh @@ -1,117 +1,196 @@ #!/bin/sh # Run this to generate all the initial makefiles, etc. -set -e +die() +{ + echo "error: $1" >&2 + exit 1 +} -srcdir=`dirname "$0"` -test -z "$srcdir" && srcdir=. +starting_point=$(pwd) -THEDIR=`pwd` -cd "$srcdir" +srcdir=$(dirname "$0") +test "$srcdir" || srcdir=. -test -f src/libvirt.c || { - echo "You must run this script in the top-level libvirt directory" - exit 1 +cd "$srcdir" || { + die "Failed to cd into $srcdir" } +test -f src/libvirt.c || { + die "$0 must live in the top-level libvirt directory" +} -EXTRA_ARGS= +dry_run= no_git= -if test "x$1" = "x--no-git"; then - no_git=" $1" - shift - case "$1 $2" in - --gnulib-srcdir=*) no_git="$no_git $1"; shift ;; - --gnulib-srcdir\ *) no_git="$no_git $1=$2"; shift; shift;; - esac -fi -if test -z "$NOCONFIGURE" ; then - if test "x$1" = "x--system"; then - shift - prefix=/usr - libdir=$prefix/lib - sysconfdir=/etc - localstatedir=/var - if [ -d /usr/lib64 ]; then - libdir=$prefix/lib64 - fi - EXTRA_ARGS="--prefix=$prefix --sysconfdir=$sysconfdir --localstatedir=$localstatedir --libdir=$libdir" - echo "Running ./configure with $EXTRA_ARGS $@" - else - if test -z "$*" && test ! -f "$THEDIR/config.status"; then - echo "I am going to run ./configure with no arguments - if you wish" - echo "to pass any to it, please specify them on the $0 command line." - fi - fi -fi +gnulib_srcdir= +extra_args= +while test "$#" -gt 0; do + case "$1" in + --dry-run) + # This variable will serve both as an indicator of the fact that + # a dry run has been requested, and to store the result of the + # dry run. It will be ultimately used as return code for the + # script: 0 means no action is necessary, 2 means that autogen.sh + # needs to be executed, and 1 is reserved for failures + dry_run=0 + shift + ;; + --no-git) + no_git=" $1" + shift + ;; + --gnulib-srcdir=*) + gnulib_srcdir=" $1" + shift + ;; + --gnulib-srcdir) + gnulib_srcdir=" $1=$2" + shift + shift + ;; + --system) + prefix=/usr + sysconfdir=/etc + localstatedir=/var + if test -d $prefix/lib64; then + libdir=$prefix/lib64 + else + libdir=$prefix/lib + fi + extra_args="--prefix=$prefix --localstatedir=$localstatedir" + extra_args="$extra_args --sysconfdir=$sysconfdir --libdir=$libdir" + shift + ;; + *) + # All remaining arguments will be passed to configure verbatim + break + ;; + esac +done +no_git="$no_git$gnulib_srcdir" -# Compute the hash we'll use to determine whether rerunning bootstrap -# is required. The first is just the SHA1 that selects a gnulib snapshot. -# The second ensures that whenever we change the set of gnulib modules used -# by this package, we rerun bootstrap to pull in the matching set of files. -# The third ensures that whenever we change the set of local gnulib diffs, -# we rerun bootstrap to pull in those diffs. -bootstrap_hash() +gnulib_hash() { + local no_git=$1 + if test "$no_git"; then - echo no-git + echo "no-git" return fi - git submodule status | sed 's/^[ +-]//;s/ .*//' + + # Compute the hash we'll use to determine whether rerunning bootstrap + # is required. The first is just the SHA1 that selects a gnulib snapshot. + # The second ensures that whenever we change the set of gnulib modules used + # by this package, we rerun bootstrap to pull in the matching set of files. + # The third ensures that whenever we change the set of local gnulib diffs, + # we rerun bootstrap to pull in those diffs. + git submodule status .gnulib | awk '{ print $1 }' git hash-object bootstrap.conf - git ls-tree -d HEAD gnulib/local | awk '{print $3}' + git ls-tree -d HEAD gnulib/local | awk '{ print $3 }' } -# Ensure that whenever we pull in a gnulib update or otherwise change to a -# different version (i.e., when switching branches), we also rerun ./bootstrap. -# Also, running 'make rpm' tends to litter the po/ directory, and some people -# like to run 'git clean -x -f po' to fix it; but only ./bootstrap regenerates -# the required file po/Makevars. -# Only run bootstrap from a git checkout, never from a tarball. +# Only look into git submodules if we're in a git checkout if test -d .git || test -f .git; then - curr_status=.git-module-status t= - if test "$no_git"; then - t=no-git - elif test -d .gnulib; then - t=$(bootstrap_hash; git diff .gnulib) + + # Check for dirty submodules + if test -z "$CLEAN_SUBMODULE"; then + for path in $(git submodule status | awk '{ print $2 }'); do + case "$(git diff "$path")" in + *-dirty*) + echo "error: $path is dirty, please investigate" >&2 + echo "set CLEAN_SUBMODULE to discard submodule changes" >&2 + exit 1 + ;; + esac + done fi - case $t:${CLEAN_SUBMODULE+set} in - *:set) ;; - *-dirty*) - echo "error: gnulib submodule is dirty, please investigate" 2>&1 - echo "set env-var CLEAN_SUBMODULE to discard gnulib changes" 2>&1 - exit 1 ;; - esac - # Keep this test in sync with cfg.mk:_update_required - if test "$t" = "$(cat $curr_status 2>/dev/null)" \ - && test -f "po/Makevars" && test -f AUTHORS; then - # good, it's up to date, all we need is autoreconf - autoreconf -if + if test "$CLEAN_SUBMODULE" && test -z "$no_git"; then + if test -z "$dry_run"; then + echo "Cleaning up submodules..." + git submodule foreach 'git clean -dfqx && git reset --hard' || { + die "Cleaning up submodules failed" + } + fi + fi + + # Update all submodules. If any of the submodules has not been + # initialized yet, it will be initialized now; moreover, any submodule + # with uncommitted changes will be returned to the expected state + echo "Updating submodules..." + git submodule update --init || { + die "Updating submodules failed" + } + + # The expected hash, eg. the one computed after the last + # successful bootstrap run, is stored on disk + state_file=.git-module-status + expected_hash=$(cat "$state_file" 2>/dev/null) + actual_hash=$(gnulib_hash "$no_git") + + if test "$actual_hash" = "$expected_hash" && \ + test -f po/Makevars && test -f AUTHORS; then + # The gnulib hash matches our expectations, and all the files + # that can only be generated through bootstrap are present: + # we just need to run autoreconf. Unless we're performing a + # dry run, of course... + if test -z "$dry_run"; then + echo "Running autoreconf..." + autoreconf -if || { + die "autoreconf failed" + } + fi else - if test -z "$no_git" && test ${CLEAN_SUBMODULE+set}; then - echo cleaning up submodules... - git submodule foreach 'git clean -dfqx && git reset --hard' + # Whenever the gnulib submodule or any of the related bits + # has been changed in some way (see gnulib_hash) we need to + # run bootstrap again. If we're performing a dry run, we + # change the return code instead to signal our caller + if test "$dry_run"; then + dry_run=2 + else + echo "Running bootstrap..." + ./bootstrap$no_git --bootstrap-sync || { + die "bootstrap failed" + } + gnulib_hash >"$state_file" fi - echo running bootstrap$no_git... - ./bootstrap$no_git --bootstrap-sync && bootstrap_hash > $curr_status \ - || { echo "Failed to bootstrap, please investigate."; exit 1; } fi fi -test -n "$NOCONFIGURE" && exit 0 +# When performing a dry run, we can stop here +test "$dry_run" && exit "$dry_run" -cd "$THEDIR" +# If asked not to run configure, we can stop here +test "$NOCONFIGURE" && exit 0 -if test "x$OBJ_DIR" != x; then - mkdir -p "$OBJ_DIR" - cd "$OBJ_DIR" +cd "$starting_point" || { + die "Failed to cd into $starting_point" +} + +if test "$OBJ_DIR"; then + mkdir -p "$OBJ_DIR" || { + die "Failed to create $OBJ_DIR" + } + cd "$OBJ_DIR" || { + die "Failed to cd into $OBJ_DIR" + } fi -if test -z "$*" && test -z "$EXTRA_ARGS" && test -f config.status; then - ./config.status --recheck +if test -z "$*" && test -z "$extra_args" && test -f config.status; then + echo "Running config.status..." + ./config.status --recheck || { + die "config.status failed" + } else - $srcdir/configure $EXTRA_ARGS "$@" -fi && { - echo - echo "Now type 'make' to compile libvirt." -} + if test -z "$*" && test -z "$extra_args"; then + echo "I am going to run configure with no arguments - if you wish" + echo "to pass any to it, please specify them on the $0 command line." + else + echo "Running configure with $extra_args $@" + fi + "$srcdir/configure" $extra_args "$@" || { + die "configure failed" + } +fi + +echo +echo "Now type 'make' to compile libvirt." diff --git a/cfg.mk b/cfg.mk index 09548fe..c04a991 100644 --- a/cfg.mk +++ b/cfg.mk @@ -1026,33 +1026,30 @@ prev_version_file = /dev/null ifneq ($(_gl-Makefile),) ifeq (0,$(MAKELEVEL)) - _curr_status = .git-module-status - # The sed filter accommodates those who check out on a commit from which - # no tag is reachable. In that case, git submodule status prints a "-" - # in column 1 and does not print a "git describe"-style string after the - # submodule name. Contrast these: - # -b653eda3ac4864de205419d9f41eec267cb89eeb .gnulib - # b653eda3ac4864de205419d9f41eec267cb89eeb .gnulib (v0.0-2286-gb653eda) - # $ cat .git-module-status - # b653eda3ac4864de205419d9f41eec267cb89eeb - # - # Keep this logic in sync with autogen.sh. - _submodule_hash = $(SED) 's/^[ +-]//;s/ .*//' - _update_required := $(shell \ - cd '$(srcdir)'; \ - test -d .git || { echo 0; exit; }; \ - test -f po/Makevars || { echo 1; exit; }; \ - test -f AUTHORS || { echo 1; exit; }; \ - test "no-git" = "$$(cat $(_curr_status))" && { echo 0; exit; }; \ - actual=$$(git submodule status | $(_submodule_hash); \ - git hash-object bootstrap.conf; \ - git ls-tree -d HEAD gnulib/local | awk '{print $$3}'; \ - git diff .gnulib); \ - stamp="$$($(_submodule_hash) $(_curr_status) 2>/dev/null)"; \ - test "$$stamp" = "$$actual"; echo $$?) + _dry_run_result := $(shell \ + cd '$(srcdir)'; \ + test -d .git || test -f .git || { echo 0; exit; }; \ + $(srcdir)/autogen.sh --dry-run >/dev/null 2>&1; \ + echo $$?; \ + ) _clean_requested = $(filter %clean,$(MAKECMDGOALS)) - ifeq (1,$(_update_required)$(_clean_requested)) - $(info INFO: gnulib update required; running ./autogen.sh first) + + # A return value of 0 means no action is required + + # A return value of 1 means a genuine error has occurred while + # performing the dry run, and it should be reported so it can + # be investigated + ifeq (1,$(_dry_run_result)) + $(info INFO: autogen.sh error, running again to show details) +maint.mk Makefile: _autogen_error + endif + + # A return value of 2 means that autogen.sh needs to be executed + # in earnest before building, probably because of gnulib updates. + # We don't run autogen.sh if the clean target has been invoked, + # though, as it would be quite pointless + ifeq (2,$(_dry_run_result)$(_clean_requested)) + $(info INFO: running autogen.sh is required, running it now...) $(shell touch $(srcdir)/AUTHORS $(srcdir)/ChangeLog) maint.mk Makefile: _autogen endif @@ -1066,6 +1063,10 @@ _autogen: $(srcdir)/autogen.sh ./config.status +.PHONY: _autogen_error +_autogen_error: + $(srcdir)/autogen.sh --dry-run + # regenerate HACKING as part of the syntax-check ifneq ($(_gl-Makefile),) syntax-check: $(top_srcdir)/HACKING spacing-check test-wrap-argv \ -- 2.7.4

2 2

[libvirt] [PATCH] test: Remove unnecessary unlocks in cleanup paths
by John Ferlan 25 Apr '17

25 Apr '17

Commit id '865f479da' altered the logic to use a common test*ObjFindByName helpers which would lock/unlock the test driver; however, a few cleanup paths in that cleanup missed removing the Unlock, so remove it now. Signed-off-by: John Ferlan <jferlan(a)redhat.com> --- src/test/test_driver.c | 5 ----- 1 file changed, 5 deletions(-) diff --git a/src/test/test_driver.c b/src/test/test_driver.c index 00999b1..2db3f7d 100644 --- a/src/test/test_driver.c +++ b/src/test/test_driver.c @@ -3931,7 +3931,6 @@ testInterfaceUndefine(virInterfacePtr iface) ret = 0; cleanup: - testDriverUnlock(privconn); return ret; } @@ -3960,7 +3959,6 @@ testInterfaceCreate(virInterfacePtr iface, cleanup: if (privinterface) virInterfaceObjUnlock(privinterface); - testDriverUnlock(privconn); return ret; } @@ -3989,7 +3987,6 @@ testInterfaceDestroy(virInterfacePtr iface, cleanup: if (privinterface) virInterfaceObjUnlock(privinterface); - testDriverUnlock(privconn); return ret; } @@ -4480,7 +4477,6 @@ testStoragePoolUndefine(virStoragePoolPtr pool) if (privpool) virStoragePoolObjUnlock(privpool); testObjectEventQueue(privconn, event); - testDriverUnlock(privconn); return ret; } @@ -4591,7 +4587,6 @@ testStoragePoolDestroy(virStoragePoolPtr pool) testObjectEventQueue(privconn, event); if (privpool) virStoragePoolObjUnlock(privpool); - testDriverUnlock(privconn); return ret; } -- 2.9.3

2 1

[libvirt] [PATCH] util: relax virNetDevSetCoalesce() stub
by Roman Bogorodskiy 24 Apr '17

24 Apr '17

Currently, virNetDevSetCoalesce() stub is always returning error. As it's used by virNetDevTapCreateInBridgePort(), it essentially breaks bridged networking if coalesce is not supported. To make it work, relax the stub to trigger error only when its coalesce argument is not NULL, otherwise report success. --- src/util/virnetdev.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 131a56a8d..131c4d983 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -3155,8 +3155,11 @@ int virNetDevSetCoalesce(const char *ifname, } # else int virNetDevSetCoalesce(const char *ifname, - virNetDevCoalescePtr coalesce ATTRIBUTE_UNUSED) + virNetDevCoalescePtr coalesce) { + if (!coalesce) + return 0; + virReportSystemError(ENOSYS, _("Cannot set coalesce info on interface '%s'"), ifname); -- 2.11.0

2 2

[libvirt] [PATCH] Fix minor typos
by Yuri Chornoivan 24 Apr '17

24 Apr '17

Hi, Attached is a minor fix for the typos in libvirt's git/master. Best regards, Yuri P. S. Sorry for the lame format. My mail provider rejects direct sending emails from git send.

2 1

[libvirt] [PATCH go-xml] Support for filesystem devices
by Ryan Goodfellow 24 Apr '17

24 Apr '17

This commit adds filesystem device support. A new family of types DomainFilesystem* are introduced and plumbed into the DomainDeviceList struct. Testing has also been included. --- domain.go | 40 ++++++++++++++++++++++++++++++++++++++++ domain_test.go | 55 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 95 insertions(+) diff --git a/domain.go b/domain.go index 307c71c..cccd9a6 100644 --- a/domain.go +++ b/domain.go @@ -81,6 +81,45 @@ type DomainDisk struct { Target *DomainDiskTarget `xml:"target"` } +type DomainFilesystemDriver struct { + Type string `xml:"type,attr"` + Name string `xml:"name,attr,omitempty"` + WRPolicy string `xml:"wrpolicy,attr,omitempty"` +} + +type DomainFilesystemSource struct { + Dir string `xml:"dir,attr,omitempty"` + File string `xml:"file,attr,omitempty"` +} + +type DomainFilesystemTarget struct { + Dir string `xml:"dir,attr"` +} + +type DomainFilesystemReadOnly struct { +} + +type DomainFilesystemSpaceHardLimit struct { + Value int `xml:",chardata"` + Unit string `xml:"unit,attr,omitempty"` +} + +type DomainFilesystemSpaceSoftLimit struct { + Value int `xml:",chardata"` + Unit string `xml:"unit,attr,omitempty"` +} + +type DomainFilesystem struct { + Type string `xml:"type,attr"` + AccessMode string `xml:"accessmode,attr"` + Driver *DomainFilesystemDriver `xml:"driver"` + Source *DomainFilesystemSource `xml:"source"` + Target *DomainFilesystemTarget `xml:"target"` + ReadOnly *DomainFilesystemReadOnly `xml:"readonly"` + SpaceHardLimit *DomainFilesystemSpaceHardLimit `xml:"space_hard_limit"` + SpaceSoftLimit *DomainFilesystemSpaceSoftLimit `xml:"space_soft_limit"` +} + type DomainInterfaceMAC struct { Address string `xml:"address,attr"` } @@ -212,6 +251,7 @@ type DomainVideo struct { type DomainDeviceList struct { Controllers []DomainController `xml:"controller"` Disks []DomainDisk `xml:"disk"` + Filesystems []DomainFilesystem `xml:"filesystem"` Interfaces []DomainInterface `xml:"interface"` Serials []DomainChardev `xml:"serial"` Consoles []DomainChardev `xml:"console"` diff --git a/domain_test.go b/domain_test.go index e5347ea..06d585c 100644 --- a/domain_test.go +++ b/domain_test.go @@ -690,6 +690,61 @@ var domainTestData = []struct { `</domain>`, }, }, + { + Object: &Domain{ + Type: "kvm", + Name: "test", + Devices: &DomainDeviceList{ + Filesystems: []DomainFilesystem{ + DomainFilesystem{ + Type: "mount", + AccessMode: "mapped", + Driver: &DomainFilesystemDriver{ + Type: "path", + WRPolicy: "immediate", + }, + Source: &DomainFilesystemSource{ + Dir: "/home/user/test", + }, + Target: &DomainFilesystemTarget{ + Dir: "user-test-mount", + }, + }, + DomainFilesystem{ + Type: "file", + AccessMode: "passthrough", + Driver: &DomainFilesystemDriver{ + Name: "loop", + Type: "raw", + }, + Source: &DomainFilesystemSource{ + File: "/home/user/test.img", + }, + Target: &DomainFilesystemTarget{ + Dir: "user-file-test-mount", + }, + }, + }, + }, + }, + Expected: []string{ + `<domain type="kvm">`, + ` <name>test</name>`, + ` <devices>`, + ` <filesystem type="mount" accessmode="mapped">`, + ` <driver type="path" wrpolicy="immediate"></driver>`, + ` <source dir="/home/user/test"></source>`, + ` <target dir="user-test-mount"></target>`, + ` </filesystem>`, + ` <filesystem type="file" accessmode="passthrough">`, + ` <driver type="raw" name="loop"></driver>`, + ` <source file="/home/user/test.img"></source>`, + ` <target dir="user-file-test-mount"></target>`, + ` </filesystem>`, + ` </devices>`, + `</domain>`, + }, + }, } func TestDomain(t *testing.T) { -- 2.11.0

2 3

[libvirt] [PATCH] [PATCH] rpc: fix keep alive timer segfault
by Yi Wang 24 Apr '17

24 Apr '17

ka maybe have been freeed in virObjectUnref, application using virKeepAliveTimer will segfault when unlock ka. We should keep ka's refs positive before using it. Signed-off-by: Yi Wang <wang.yi59(a)zte.com.cn> --- src/rpc/virkeepalive.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) diff --git a/src/rpc/virkeepalive.c b/src/rpc/virkeepalive.c index c9faf88..4f666fd 100644 --- a/src/rpc/virkeepalive.c +++ b/src/rpc/virkeepalive.c @@ -160,17 +160,17 @@ virKeepAliveTimer(int timer ATTRIBUTE_UNUSED, void *opaque) bool dead; void *client; + virObjectRef(ka); virObjectLock(ka); client = ka->client; dead = virKeepAliveTimerInternal(ka, &msg); + virObjectUnlock(ka); + if (!dead && !msg) goto cleanup; - virObjectRef(ka); - virObjectUnlock(ka); - if (dead) { ka->deadCB(client); } else if (ka->sendCB(client, msg) < 0) { @@ -178,11 +178,8 @@ virKeepAliveTimer(int timer ATTRIBUTE_UNUSED, void *opaque) virNetMessageFree(msg); } - virObjectLock(ka); - virObjectUnref(ka); - cleanup: - virObjectUnlock(ka); + virObjectUnref(ka); } -- 1.8.3.1

4 5

[libvirt] [PATCH v2 0/2] Fix possible use-after-free when sending event message
by John Ferlan 24 Apr '17

24 Apr '17

v1: https://www.redhat.com/archives/libvir-list/2017-March/msg01228.html Change since v1, add the derefFcn as an argument to the renamed macro (not quite sure how I missed that originally. John Ferlan (2): daemon: Rework remoteClientFreeFunc cleanup loops into C macro remote: Fix possible use-after-free when sending event message daemon/remote.c | 164 ++++++++++++++++++++------------------------------------ 1 file changed, 58 insertions(+), 106 deletions(-) -- 2.9.3

3 7

[libvirt] [PATCH] util: fix virNetDevSetCoalesce fallback on Win32/FreeBSD
by Daniel P. Berrange 23 Apr '17

23 Apr '17

The current fallback stub for virNetDevSetCoalesce is inside an earlier conditional block. This deals with the feature being missing on older Linux platforms. We need a second fallback stub though, outside the top level conditional, to ensure builds work on Win32/FreeBSD platforms too. Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com> --- Pushed as a build fix src/util/virnetdev.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/src/util/virnetdev.c b/src/util/virnetdev.c index 6ff1b48..27f1637 100644 --- a/src/util/virnetdev.c +++ b/src/util/virnetdev.c @@ -3207,6 +3207,15 @@ virNetDevGetFeatures(const char *ifname ATTRIBUTE_UNUSED, ifname); return 0; } + +int virNetDevSetCoalesce(const char *ifname, + virNetDevCoalescePtr coalesce ATTRIBUTE_UNUSED) +{ + virReportSystemError(ENOSYS, + _("Cannot set coalesce info on interface '%s'"), + ifname); + return -1; +} #endif -- 2.9.3

3 3

[libvirt] [PATCH] Remove pointless check for !ret in virDomainNetDefCoalesceParseXML
by Martin Kletzander 22 Apr '17

22 Apr '17

It was left there after removing a macro it was part of in first version or so. Now it will always be NULL. Signed-off-by: Martin Kletzander <mkletzan(a)redhat.com> --- Pushed as trivial src/conf/domain_conf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/conf/domain_conf.c b/src/conf/domain_conf.c index 4a777d67bfad..7d3cf178eaf6 100644 --- a/src/conf/domain_conf.c +++ b/src/conf/domain_conf.c @@ -6789,7 +6789,7 @@ virDomainNetDefCoalesceParseXML(xmlNodePtr node, if (!str) goto cleanup; - if (!ret && VIR_ALLOC(ret) < 0) + if (VIR_ALLOC(ret) < 0) goto cleanup; if (virStrToLong_ullp(str, NULL, 10, &tmp) < 0) { -- 2.12.2

1 0